ABSTRACT Objective To investigate the value of Automated breast volume scanning (ABVS) ultrasonographic features in preoperative prediction of axillary lymph node metastatic burden. Methods A retrospective analysis of 106 cases of breast cancer

which was confirmed by surgical pathology of ABVS ultrasonographic features

according to the burden of axillary lymph node metastasis

points of low burden group (≤2 metastasis lymph nodes) and high burden group (≥3 metastasis lymph nodes). The differences in ultrasonic features of the two groups were compared and studied. The factors with statistical significance in univariate analysis were included into the binary Logistic regression model and the regression equation was established. Receiver operating characteristic (ROC) curve was drawn and the Area under curve (AUC) was calculated to analyze the diagnostic efficiency of the model. Results Single factor analysis of ultrasonic features between the low burden group and the high burden group showed no statistical significance in the location

morphology

growth orientation

internal echo

posterior echo

calcification and coronal high-echo halo (P＞0.05). There were statistically significant differences in the maximum diameter of lesion

marginal condition

distance to nipple

distance to skin

retraction phenomenon on the coronal planes

worm biting sign on the coronal planes

and blood flow signal (P＜0.05). Multi-factor binary Logistic regression analysis was used to analyze the maximum diameter (OR=4.971

P=0.007)

distance to skin (OR=3.559

P=0.017)

retraction phenomenon on the coronal planes (OR=5.932

P=0.019)

worm biting sign on the coronal planes (OR=9.426

P=0.003) and blood flow signal (OR=3.367

P=0.033) were independent risk factors for high burden of axillary lymph node metastasis. Logistic regression equation was as follows: Logistic (P) =-4.402+1.604× maximum diameter ＞2 cm+1.270× distance to skin ≤0.2 cm+1.780× retraction phenomenon +2.244× entomophagy phenomenon +1.214× blood flow signal Ⅱ~Ⅲ. The Logistic regression model demonstrated that with prediction probablity P=0.50 as the cut-off value

sensitivity was 71.7%，specificity was 81.7%，the diagnostic accuracy was 77.4% and area under ROC curve(AUC) was 0.872. Conclusion The model of axillary lymph node metastatic burden prediction based on multivariate Logistic regression analysis has a certain predictive effect on high lymph node metastatic burden. It has a higher diagnostic accuracy and a certain clinical practical value.